Quaternion Rotation Biradial - Wikipedia, The Free Encyclopedia20150415003245

8/15/2019 Quaternion Rotation Biradial - Wikipedia, The Free Encyclopedia20150415003245

1/61

Biradial b/a

Quaternion rotation biradialFrom Wikipedia, the free encyclopedia

In mathematics, a quaternion biradial[1](Art.93) is the quotient (or product , , , ) of two pure quaternion

vectors and , sometimes called rays.

Consider the quaternion biradial (read “b by a”). The biradial is an operator

that turns into as an operator on : . The biradial turningoperation is a composition of a rotation and a scaling that rotates into the line of ,

followed with scaling by . If then and acts as only a rotation

operator (a.k.a., a versor , rotor , or 2D spinor ) in the -plane through the angle from

to .

Contents

1 Quaternion spatial rotations explained using Hamilton’s biradial concept

1.1 Quaternion identities

1.2 The biradial b/a

1.3 Quadrantal versors and handedness

1.4 Quaternions and the left-hand rule on left-handed axes

1.5 Rotation formulas

1.6 The scalar power of a vector

1.7 Rotation of vector components parallel and perpendicular to a rotation

axis

1.8 The product ba

1.9 The biradial a/b1.10 The product ab

1.11 Pairs of conjugate versors

1.12 The rotation formula (b/a)v(a/b)

1.13 The rays a,b of biradial b/a need not be unit vectors

1.14 The conical rotation by pi, reflection, or conjugation ba/b

1.15 Rotations as successive reflections in vectors

1.16 Equivalent versors

1.17 The square root of a versor

1.18 Rodrigues rotation formula and identity to quaternion rotation formula

1.19 Rotation round arbitrary lines and points

1.20 Vector-arcs

1.21 Mean versor

1.22 Matrix representations of quaternions and rotations

1.23 Relationship between quaternions and complex numbers

2 Quaternion rotation biradials in Clifford geometric algebra

2.1 Quaternion vectors are imaginary directed quantities

2.2 Euclidean vectors are real directed quantities

2.3 Clifford geometric algebras

2.4 Rules for the quaternion and Euclidean vector units

2.5 The outer product

ternion rotation biradial - Wikipedia, the free encyclopedia https://en.wikipedia.org/w/index.php?title=Quaternion_rotation_biradi...

d 61 15.4.2015 0:32


2/61

2.6 Outer products are geometrical values for imaginaries

2.7 Outer product is identified with scalar multiplication

2.8 Grades, blades, and vectors

2.9 The inner product

2.10 The sum of the inner and outer products

2.11 The reverse of a blade

2.12 The square of a blade

2.13 The geometric product ab of Euclidean vectors

2.14 The quaternion product PQ of even-grade multivectors

2.15 Dual and undual to convert to and from quaternions

2.16 The inner product of quaternionic bivectors has negative sign

2.17 The commutator product

2.18 Evaluating expressions in Clifford geometric algebra

2.19 The general form of a multivector A

2.20 The algebra of the even-grade parts is quaternions

2.21 Representing a quaternion versor using an odd-grade part

2.22 The rotation (b/a)v(a/b) in Euclidean vectors

2.23 Identities: Euclidean base unit vectors

2.24 Identities: Euclidean 1-vectors

2.25 Identities: The geometric product ba of 1-vectors

2.26 Identities: The quaternion unit 2-vectors

2.27 Quaternions mapped into a two dimensional space

2.28 Identities: General form of multivector A by grade-k parts

2.29 Identities: Squares, inverses, and commutative property of I

2.30 Identities: Product ba and rotation operator R

2.31 Identities: Vector multivector products aB and Ab

2.32 Operator precedence convention for ambiguous products

2.33 Identities: Projection and rejection operators for vectors

2.34 Associative, distributive algebraic properties of products

2.35 Recursive formulas for the inner products of blades

2.36 Expansion of the geometric product of blades

2.37 Definitions of products by graded parts of geometric product

2.38 Identities: Inner products2.39 Triple vector cross product

2.40 Rotation of blades

3 References

Quaternion spatial rotations explained using Hamilton’s biradial concept

Quaternion identities

For simplicity in the following explanations, let so that and are unit vectors. Frequent use will bemade of the following identities.

Definition of quaternion units and products:


d 61 15.4.2015 0:32


3/61

Unit vectors:

Conjugate, tensor, and versor operations:

Inverses of a quaternion and a vector:

Vector products:

The operations , , , , are the scalar part, vector part, conjugate, tensor , and versor of quaternion in

Hamilton’s notation and terminology.

Since scalars commute with all elements in the quaternion algebra, so do scalars such as , , , and .

For perpendicular vectors , and . Therefore, , , and .

For any scalar and value where , the Taylor series for gives the Euler formula


d 61 15.4.2015 0:32


4/61

The biradial b/a

The biradial can be written

where is the component of parallel to , is the component of perpendicular to , is the angle from to ,

and is the unit vector normal to the -plane.

Quadrantal versors and handedness

When then , showing that is the quadrantal versor of the -plane perpendicular to . We can also

write

where is the number of quadrants (multiples of radians) of rotation round as the axis of rotation following a

right-hand rule on a right-handed axes model or following a left-hand rule on a left-handed axes model. If is at a positive

angle in the plane from , this means that on right-handed axes, rotates by right-hand rule counter-clockwise from

into , while on left-handed axes rotates by left-hand rule clockwise from into . Angles are positive counter-

clockwise on right-handed axes, and are positive clockwise on left-handed axes. The choice of right-handed or left-handed

axes is a modeling choice outside the algebra that affects the geometric interpretation of the algebra but not the algebra

itself.

The relation between the unit vectors as defined by Hamilton, also defines as the

quadrantal versors or biradials

following a right-hand (counter-clockwise) rule on right-handed axes or a left-hand (clockwise) rule on left-handed axes,

where each rotate by radians in their perpendicular plane according to the matching handedness rule and axesmodel being used.

Quaternions and the left-hand rule on left-handed axes

In a sense, a left-hand rule and axes are a better fit to Hamilton’s defined relation since then a versor used as left-side

multiplier to operate on a right-side multiplicand performs left-hand clockwise rotation for positive angles, and the same

versor used as a right-side multiplier to operate on a left-side multiplicand performs right-hand counter-clockwise rotation

by the same angle (the reverse rotation). For example, imagine working on left-handed axes and interpret the geometry

represented by the following quadrantal versor rotations


d 61 15.4.2015 0:32


5/61

which show that operating with on the left-side performs left-hand clockwise rotation, operating with on the right-side

performs a right-hand counter-clockwise (reverse) rotation, and operating with on the right-side performs anotherclockwise rotation.

On left-handed axes, there is arguably more consistency of handedness as if the quaternion relation

was defined for a left-handed system of axes. It is also possible for the relation to have

been defined , inverting the quaternion space and making right-handed axes arguably match

the geometric interpretation of the algebra better. In Hamilton’s book Lectures on Quaternions[1] he often uses

left-handed[1](Art.228) axes with left-hand rule clockwise rotations, and it seems that was often the preferred orientation used

by the inventor of quaternions.

Rotation formulas

These results so far will be shown to generalize for any quadrantal versor rotating a vector by quadrants or

by radians in the plane perpendicular to as

where for quadrant the result is the reflection of in , or conically rotated round , or is conjugated by .

Note that we need to convert angles in quadrants to radians since the ordinary trigonometric functions , are

defined for in radians.

The scalar power of a vector

Any versor can be constructed or expressed using its rotational angle in quadrants and its unit vector or quadrantal versor

rotational axis as . This expression seems to be seldomly used but is important to understand since it gives a clear

geometric meaning to taking the scalar power of a vector. For a general vector , its scalar power of can be

expressed as and still represents a versor that rotates round by quadrants or radians,

and also a scaling or tensor (in Hamilton’s terminology and notation) that is applied times.

Rotation of vector components parallel and perpendicular to a rotation axis

When is a vector in the plane perpendicular to , then and are perpendicular and obey the anti-commutativemultiplication property , leading to

which demonstrates that the conjugate versor applied as right-side operator on rotates the same (in the plane

perpendicular to the versor axis ) as the versor itself applied as left-side operator on , and vice versa. The reverse

rotation can be written

which is also seen as two successive rotations with the second outer rotation undoing or reversing the first inner rotation.

When is a vector parallel to , then their multiplication is commutative, leading to


d 61 15.4.2015 0:32


6/61

where is unaffected by the rotation.

In general, a vector will be neither completely parallel nor perpendicular to another vector . Similar to how was

written above in the expressions for , a vector can be written since it is always possible for any

vector to be written as a sum of components parallel and perpendicular to any other vector . The parallel part

is the projection of on . The perpendicular part

is the rejection of by . We can also write the orthogonalization of by as

Combining some results from above we get

which represents the component unaffected by a null (or any) rotation round , while the component undergoes

a planar rotation by in the plane perpendicular to . The sum of these vectors is the rotated version of . This is an

important formula for the type of rotation known as conical rotation by round the line of relative to the origin. The

cone apex is the origin, the cone base is in the plane perpendicular to and intersecting at distance from the

origin, and the sides of the cone are swept by the line of as it is rotated round axis for a complete rotation.

A versor and its conjugate have the same angle but inverse axes

or they have the same axis but reverse angles

however, there is often a preference to take positive angles and inverse axes. The conjugate or inverse versor is

sometimes also called the reversor of .

We’ve looked at the biradial already, so let’s now look at the other products.

The product ba

The product gives


d 61 15.4.2015 0:32


7/61

showing that, compared to the quotient , the product rotates by an additional angle or an angle round

the same axis , or rotates by the supplement angle in the opposite direction round the inverse axis . Notice

how multiplying by is the same as rotating by an additional round the same axis.

The biradial a/b

The biradial gives

which is the inverse of , as expected, that rotates by round the same axis .

The product ab

The product gives

showing that, compared to the quotient , the product rotates by the supplement angle round the same axis

.

Pairs of conjugate versors

Notice that & , and & are two pairs of conjugateversors. Conjugates have the same angle of rotation but round inverse axes, and therefore rotate in opposite directions. The

conjugate of a versor is also its inverse, so they are both pairs of inverses.

Conjugate versors: products and quotients of unit vectors.

The rotation formula (b/a)v(a/b)

For the -plane of the biradial , the unit vectors and span the plane and are a basis for vectors in the plane so

that any vector in the plane can be written in the form . The unit vector spans the 3rd linearly

independent spatial dimension. An arbitrary vector of the pure quaternion vector space can be written


d 61 15.4.2015 0:32


8/61

, where is the component in the -plane and is the component

perpendicular to the -plane.

The biradial applied to an arbitrary vector gives

where is the component of in the -plane and rotated by in the plane. The other value

is another quaternion that rotates by and scales by , and by itself it is not a useful part of the result;

this quaternion value can be eliminated as follows

where we recognize . This suggests trying the following

and finding that, in general, rotates the component of in the plane of the biradial by twice the angle of the

biradial, and it leaves the component of perpendicular to the plane unaffected. This type of rotation is called conical

rotation due to how the line turns on a cone as it is rotated through an entire round as the axis of the cone.

The rays a,b of biradial b/a need not be unit vectors

Although and were treated as unit vectors for simplicity, the result

shows that and need not be unit vectors. Since the lengths of and are inconsequential in the rotation, they are

sometimes called rays of the biradial instead of vectors, as mentioned at the beginning.

The conical rotation by pi, reflection, or conjugation ba/b

In the work above, we encountered as a planar rotation; this can also be viewed as a conical rotation

by round as the cone axis, or viewed as a reflection of in


d 61 15.4.2015 0:32


9/61

Rhombus bisector a+b.

This result is also called the conjugation of by .

For quadrantal versor (unit vector) perpendicular to vector ,

which is the conjugation of or , and it is also called the inversion of .

Rotations as successive reflections in vectors

The conical rotation of by in the plane of the biradial with angle can also be seen as two successive

reflections

where is the first reflection of in , and is the second reflection of in .

Equivalent versors

Every pair of unit vectors in the same plane and having the same angle from to form the same biradial

with same angle and same normal

The square root of a versor

Since rotates by , if we want to rotate by just then we need to take the square root of as

resulting in the formula for rotation by in the plane normal to unit vector

In theorems it is proved that any quaternion rotation can be expressed as successive reflections in vectors that reduces

to a single conical rotation round a single resultant versor axis by a certain angle , which may be expressed as the versor

operator .

To apply a rotation by in steps, use the th-root of , and its conjugate, successively times

If , then the versor can also be written as


d 61 15.4.2015 0:32


10/61

and the rotation of vector by in the -plane, or round its normal , can be written as

so that the rotation is seen as various reflections. For example, if , then

and this rotation is the reflection of in .

If , then the sum is the diagonal vector that bisects the rhombus parallelogram framed by and and is

the reflector of into or into .

Rodrigues rotation formula and identity to quaternion rotation formula

Another formula for rotation can be derived without quaternions, using only vector analysis which includes vector addition

and subtraction and the vector dot and cross products. The formula is known as the Rodrigues Rotation Formula

which rotates any vector conically round the unit vector axis by the angle radians counter-clockwise on right-handed

axes or clockwise on left-handed axes according to the matching right-hand or left-hand rule.

This Rodrigues formula is a known result that can be derived without too much difficulty by modeling the problem in

vector analysis and finding the solution for . This formula can also be found from the quaternion rotation formula

derived already in terms of biradials

The identity of the quaternion rotation formula to the Rodrigues rotation formula is


d 61 15.4.2015 0:32


11/61

where again the unit vector is the axis of conical rotation of vector by according to the handedness described above.

Rotation round arbitrary lines and points

By proper choice of the rotation axis and the angle , the conical rotation can rotate to any location on the

sphere having center and radius . These conical rotations of occur in circles

located on the surface of the sphere, which we can call the sphere of .

There are two types of these circles. A small circle has a radius , and a great circle has radius . If isperpendicular to , the circle of rotation on the sphere is a great circle, and otherwise a small circle. The interior plane of a

circle is the base of a rotational cone, with tip or apex at the origin . As is rotated, the line of moves on the sides of

the cone, and points to different points on the circle at the cone base.

The line of the rotation axis intersects and intersects the centers of a group of parallel circles in space that section

the sphere of by cuts perpendicular to . The line of also intersects the sphere of at the two points called the poles

of the rotation where the small circles vanish or degenerate into the poles. The great circle, on the sphere of cut

perpendicular to through the center of the sphere of (always the origin of a rotation ), contains the points called

the polars of the rotation .

When is one of the two poles (a pole vector ) of the rotation , then is parallel to and it rotates in the point

(degenerated circle) of the pole and is unaffected by the rotation . When is a polar (a polar vector ) of the rotation

, then is perpendicular to and is rotated in the great circle of the sphere of perpendicular to , polar to pole

. Otherwise, the rotation of can be viewed in terms of its components parallel and perpendicular to the axis as a

cylindrical rotation

where is polar of the rotation (planar rotation) and is a pole of the rotation (null rotation). The

component is rotated in the base of the cylinder in the plane perpendicular to intersecting the origin of the

rotation, and the component is the translation up the side of the cylinder to the small circle on the sphere

of . The vector component is also called an eigenvector (“eigen” is German for “own”) of the rotation . The

sum is still rotated on a cone inside the cylinder and is a conical rotation of round .

The quadrantal versor, unit vector, could also be called an axial vector or a pole vector of a rotation on the sphere of ,

the unit sphere. All vectors perpendicular to the axial vector could be called planar-polar vectors relative to all

vectors parallel to which could be called axial-pole vectors, where could also be called the unit axial-pole

vector . Any circle round axis , in a plane perpendicular to and centered on the line of , could be called a polar circle

of . The rotation of a planar-polar vector of the rotation is a planar rotation

and the rotation of an axial-pole vector (eigenvector) of the rotation is a null rotation

The pole round which counter-clockwise rotation is seen by right-hand rule on right-handed axes (holding the pole vector in

right-hand) can be called the north geometric pole of the rotation, and the inverse pole called the south geometric pole of

the rotation where left-hand rule on left-handed axes would see clockwise rotation (holding the inverse pole vector in

left-hand as if it were the positive direction of the axis). A pole, as a point on the sphere, or as a vector from the sphere

center to the pole, are synonymous for the purpose considered here, but they can be made as distinct mathematical values

in the more advanced Clifford geometric algebras of generalized projective geometry. With this terminology, the quadrantal

versor is a unit axial-pole vector of the unit sphere of (the unit sphere) and is directed toward the point of the north

geometric pole of the rotation on right-handed axes, or directed toward the south geometric pole on left-handed axes wherepositive rotations are then taken clockwise. This convention or orientation of north and south geometric poles matches the

geometric poles and rotation of the Earth.

If the sphere of relative to an origin is considered the set


d 61 15.4.2015 0:32


12/61


13/61

as to and . The vector-arcs, or chords, of the rotations correspond. Some successive rotations

applied to produces successive vector-arcs . The vector-arcs can be added successively to vector

as and each one of these vectors remains on a sphere or path corresponding to the sphere or path of

.

Since the order, or sequence, of adding vector-arcs onto (adding successively) is important for replaying a sequence of

vector-arc translations on a sphere or path, when a vector-arc is added to , the next vector-arc added to

is a specific vector-arc representing the next rotation and is called the provector-arc of the prior vector-arc . The

sum of a vector-arc and its provector-arc is their transvector-arc representing an arcual sum[1]

(Art.218).

The addition of vector-arcs equals successive rotations in the same sphere only if the vector-arcs are successive

corresponding chords in the same sphere. The successive chords form a gauche[1](Art.323) or skew polygon inscribed in the

sphere. To remain on the same sphere, the addition of the vector-arcs to must be in rotations order, translating across a

specific sequence of chords representing the rotation arcs and sides of a gauche polygon inscribed in the sphere. As chords

are added to in rotations sequence, the resulting positions of vertices of the polygon in the sphere are found. The effect is

to replay a specific sequence of translations starting at , with vector-arcs added head to tail as the polygon sides or chords

in the sphere. Direct rotations among the found points of the inscribed polygon can be computed as other transvector-arcs.

A sequence of vector-arcs can be pre-generated and applied in sequence forward (adding the provector-arc) and backward

(subtracting the prior provector-arc). The sequences of vector-arcs could be applied to multiple points using additions ortheir pre-additions (transvector-arcs), which may be much faster than multiplying each point using a quaternion operator.

Mean versor

The mean versor problem can be described as follows: If given versors

where each versor rotates by the angle in the plane perpendicular to the unit vector , then the vector can be rotated

into different points using the half-angle (or full-angle) versors

For , the three vectors are the vertices of a spherical triangle . The location of the spherical centroid of

represents the true mean rotation of relative to the three vertices.

As a first approach to trying to find and the mean versor , we can consider the average of the vectors , scaled to

touch the sphere, as approximately . With this approach, the spherical centroid of is approximately

on the condition that the vertices are clustered closely together such that the angle between any two vertices does not

exceed radians. For any , this condition is satisfied if

when using the half-angle versors, or if

when using the full-angle versors.

Under this condition, this formula for the spherical centroid of has good accuracy. For angles exceeding these conditions,

the error increases rapidly up to the error extremum or worse-possible error when two vertices reach radians apart!

For , the vector is generally not the centroid of a spherical polygon of the vertices , but does approximate the

true mean rotation of .


d 61 15.4.2015 0:32


14/61

The approximate mean versor is

The approximate mean versor with half the angle is

The rotation of by the approximate mean versor is

Using the identities

we can verify that

By this first approach, this approximate mean versor is dependent on a particular and adapts well to a wide range

versors . It is computationally expensive since it requires computing all the rotations and averaging them. A potentialproblem of this approach is the possibly unacceptable error when two different are very close to radians apart, or are

nearly inversions.

As a second approach to trying to find and the mean versor , we can consider using the average vector-arc of the

vector-arcs represented by the versors . By this approach, we try the geometric mean of the versors

as approximating the true mean versor. Each can be interpreted as a vector-arc, and

as the average vector-arc.

By this method, the approximate mean rotation of is

Good accuracy is achieved using the same condition as in the first approach.

As a third approach to trying to find and the mean versor , the solution is to find the versor that minimizes the

distance , where


d 61 15.4.2015 0:32


15/61

We can also choose to minimize the sum of squares

We can write as

Now we can write as

If

then and is minimized. This can be expanded as

which shows more clearly that the closer is to each , then the closer is to zero. This minimization is achieved by

the arithmetic mean

This result does not depend on a particular . Technically, a versor should always have unit magnitude, but generally

. The quaternion can be normalized to have unit magnitude if necessary, but when used in the following versor

rotation operation, any magnitude of cancels anyway.

By this approach, the approximate mean rotation of is

Once again, good accuracy is achieved using the same condition as in the first and second approaches.

Each approach has a different behavior as angles are increased. For , they give nearly identical results. Each

approach may be worth comparing to the others for a particular problem.

Testing the approaches: One way to experiment interactively with these formulas is to use the program GAViewer

(http://geometricalgebra.org/gaviewer_download.html) and have it open and run the following file mean-versor.g:

/* mean-versor.g Mean Versor Demo */

// Use ctrl-rightmousebutton to drag vectors v r1 r2 r3

// around the sphere and see rotations change dynamically.


d 61 15.4.2015 0:32


16/61

// Sliders for angles t1 t2 t3 also trigger dynamic changes.

// Lines ending with , ARE rendered

// Lines ending with ; ARE NOT rendered

// Reset the viewer.

reset();

set_window_title("Mean Versor Demo");

// Run a .geo format command.

// Certain things can only be done with geo commands.

// Set the text parsing/formatting mode for the text of labels.

cmd("tsmode equation");

// Equation mode formats text like Latex equation mode does.

// Draw basic x,y,z direction unit vectors.

x=e1,

y=e2,

z=e3,

label(x),

label(y),

label(z),

// Pseudoscalar I.

// Needed to compute duals; i.e,

// to convert vectors to pure quaternions in geometric algebra.

I = e1 e2 e3;

// Draw a unit ball or sphere.

B = 4/3 pi pow(0.80,2) e1 e2 e3;

B = color(B,1,1,1,0.25),

// Initial point v on unit radius sphere.

// This is the vector we are rotating.v = (e1+e2+e3);

v = v / norm(v);

label(v),

// Make angles t1,t2,t3 in degrees here

// but converted to radians in formulas later.

ctrl_range(t1= 40,-180,180);

ctrl_range(t2=-25,-180,180);

ctrl_range(t3=-15,-180,180);

// Use half-angle versor rotation formulas below.

// Make versors R1 R2 R3 that rotate round axes r1 r2 r3

// that are initially close to e1 e2 e3 (but can be dragged).

r1 = (5 e1 + e2 + e3);

r2 = ( e1 + 5 e2 + e3);

r3 = ( e1 + e2 + 5 e3);

r1 = r1 / norm(r1),

r2 = r2 / norm(r2),

r3 = r3 / norm(r3),label(r1),

label(r2),

label(r3),

R1 = exp( t1/2 pi/180 r1/I );

R2 = exp( t2/2 pi/180 r2/I );

R3 = exp( t3/2 pi/180 r3/I );

// Rotate v into v1 v2 v3 using R1 R2 R3.

v1 = R1 v / R1,

v2 = R2 v / R2,

v3 = R3 v / R3,

label(v1),

label(v2),

label(v3),

// Locate the mean rotation of v directly from the mean of v1 v2 v3.

// This is the accurate approximation of the centroid or mean rotation of v.

// This is dependent on a particular v and on the versors R1 R2 R3.

m = (v1+v2+v3)/norm(v1+v2+v3),label(m),

// Locate the mean rotation of v using the geometric mean of R1 R2 R3

// which is thought of as using the mean vector-arc.

G = exp( ((t1/2 pi/180 r1)+(t2/2 pi/180 r2)+(t3/2 pi/180 r3))/(3I) );

g = G v / G,

label(g),

// Locate the mean rotation of v using the arithmetic mean of R1 R2 R3

// normalized to unit norm for a proper versor.

A = (R1+R2+R3)/norm(R1+R2+R3);

a = A v / A,

label(a),

// Begin dynamic element.

// Simply recalculate everything in here.

dynamic{d1:

// The basis should not be moved.

x=e1,

y=e2,

z=e3,

x= color(x,0,0,0,0.5),

y= color(y,0,0,0,0.5),

z= color(z,0,0,0,0.5),


d 61 15.4.2015 0:32


17/61

// The sphere.

B = 4/3 pi pow(0.80,2) e1 e2 e3;

B = color(B,1,1,1,0.25),

// The vector we are rotating around a unit radius sphere.

v = v / norm(v),

// Make versors R1 R2 R3 that rotate round axes r1 r2 r3

// and that are initially set close to e1 e2 e3.

r1 = r1 / norm(r1),

r2 = r2 / norm(r2),

r3 = r3 / norm(r3),

r1 = color(r1,1,1,0,0.5),

r2 = color(r2,1,1,0,0.5),

r3 = color(r3,1,1,0,0.5),

R1 = exp( t1/2 pi/180 r1/I );

R2 = exp( t2/2 pi/180 r2/I );

R3 = exp( t3/2 pi/180 r3/I );

// Rotate v into v1 v2 v3 using R1 R2 R3.

v1 = R1 v / R1,

v2 = R2 v / R2,

v3 = R3 v / R3,

v1 = color(v1,0,0,1,0.5),

v2 = color(v2,0,0,1,0.5),

v3 = color(v3,0,0,1,0.5),

// Locate the mean rotation of v directly from the mean of v1 v2 v3.

m = norm(v) (v1+v2+v3)/norm(v1+v2+v3),

m = color(m,1,0,1,0.5),

// Locate the mean rotation of v using the geometric mean of R1 R2 R3.G = exp( ((t1/2 pi/180 r1)+(t2/2 pi/180 r2)+(t3/2 pi/180 r3))/(3I) );

g = G v / G,

g = color(g,1,0,1,0.5),

// Locate the mean rotation of v using the arithmetic mean of R1 R2 R3.

// The magnitude of A does not matter here.

A = (R1+R2+R3);

a = A v / A,

a = color(a,1,0,1,0.5),

}

Matrix representations of quaternions and rotations

A quaternion can be written as a -tuple of scalar components or as a -element column vector, and it can also be written

as a matrix by expanding the quaternion product

and factoring the resulting column vector form into a matrix-vector product .

Quaternions in the matrix form can be added as , and can be multiplied together

using non-commutative matrix multiplication giving products of the form . Quaternions in


d 61 15.4.2015 0:32


18/61

column vector form can be added, and can be the multiplicand of a quaternion multiplier in matrix form ,

giving products in column vector form.

For two vectors, their product is

and their matrix-vector product is

which gives matrices representing dot and cross product operators on column vectors.

The dot and cross products for two vectors in matrix forms are

which take the symmetric and anti-symmetric parts of the vector product .

The matrix form can be decomposed into a sum of matrices to identify the matrix representations of as


d 61 15.4.2015 0:32


19/61

A quaternion rotation, by radians, of vector round the unit vector axis can be written as

To write this rotation in matrix-vector multiplication form, we first need some more identities.

The vector component of parallel to the unit vector is

The vector component of perpendicular to the unit vector is


d 61 15.4.2015 0:32


20/61

The rotor for the rotation round axis by angle is

The rotation can also be written

which appears to be a smaller form of the Rodrigues rotation formula. Starting here, we have enough identities to switch

into matrices and column vectors as

which can be written as the matrix-vector product


d 61 15.4.2015 0:32


21/61

where this matrix represents the rotation operator. The lower part of the matrix is all that is needed to rotate vectors

in 3D. The full matrix can be used to rotate quaternions, where any scalar part is unaffected by the rotation and the

vector part is rotated in the usual way.

Rotations round the axes are found by setting as one of them, which gives

A rotation by a negative angle is the same as a change of basis onto axes rotated by the positive angle. Cramer's rule, and

related methods, provide a more generalized way to compute vectors relative to a new or changed basis.

Relationship between quaternions and complex numbers

The relationship between complex numbers and quaternions, with respect to planar rotations by radians in the vector

plane perpendicular to , might be written as

This seems to show that must be identified with representing a planar rotation by or a null rotation round ,

otherwise the results degenerate. The value (not to be confounded as the quaternion unit ) seems to represent rotation by

round since

In general, seems to represent quadrants of rotation round . However, this interpretation of holds for any arbitrary

axis , such that can always be interpreted as a versor that is perpendicular to any and all geometrical (generally

non-commutative multiplication) planes spanned by arbitrary biradials, or that is parallel and equivalent to all axes of

rotation on the unit-sphere.


d 61 15.4.2015 0:32


22/61

On a metrical (generally commutative multiplication) complex number plane where points are represented by

complex numbers , the value is a quadrantal rotation operator on the points

If is temporarily identified with (as an isomorphism), then can be interpreted simultaneously as both a

geometrical rotation operator on geometrical vectors in the -plane and as a metrical rotation operator on

metrical points represented by complex numbers .

In standard quaternions, the scalars are limited to real numbers so that complex numbers are not

elements of the set of standard quaternion numbers. This means that using within standard or real quaternions is not

permitted. Therefore, the identification of with , or trying is not valid. In the above, the words “might” and

“seems” are emphasized because , or any power of , is not a valid value to construct within standard

quaternions. Stated more clearly, the power

is generally a complex number, and therefore it is not generally permitted in real quaternions.

Whenever a value similar to is needed, a specific versor having specific geometrical significance on theunit-sphere must be used as a metrical rotation operator on the -plane of pseudo-complex numbers .

In practice, an exception can be made for and allow this power if is an integer, since the result remains a real

number. Expressions of the form are often used to give the sign or orientation of certain results. In general, this

exception can cause problems because it is an operation outside of the proper algebra of quaternions.

Consider the general case of taking the power of a quaternion

and notice that tensors are positive and the power of a versor should never require taking the power so long as the

axis is known.

If or then and can be lost, which is a problem where it can be said that a scalar , or any

scalar by itself, is not a quaternion. It is important to not lose the axis of the quaternion or else the quaternion geometric

algebra is not closed, or it has exited into the metrical algebra of real or complex numbers if that is acceptable to your

application. Whatever the angle of , the versor expression should be retained and not allowed to degenerate into a

scalar if remaining in the quaternion algebra is important.

It would appear that quaternions do not really extend metrical real or complex numbers, but forms its very own specialgeometric algebra that is not closed if the axis (the geometric part) of a quaternion is lost. The paper Planar Complex

Numbers in Even n Dimensions and the monograph Complex Numbers in n Dimensions, both by Silviu Olariu and

published in 2000, present a possible extension of complex numbers that have commutative multiplication. Higher

dimensional extensions of complex numbers are also known as hypercomplex numbers.

In geometric algebras, such as quaternions and Clifford geometric algebras, the value is usually avoided due to its

ambiguous or indeterminate geometrical interpretation. It is avoided by limiting scalars to real numbers and using a

subalgebra isomorphic to complex numbers when needed. Again, it is also possible to allow the geometric algebra to exit

into the metrical real or complex number algebra if that is acceptable. And again, an exception is integer powers

that are commonly used in formulas to compute signs or orientation.

The biquaternions[1](Art.669) include as an element of its algebra, but it might be well to also consider using the Clifford

geometric algebra , wherein complex numbers, quaternions, biquaternions, and dual quaternions are all embedded

isomorphically as subalgebras.


d 61 15.4.2015 0:32


23/61

Quaternion rotation biradials in Clifford geometric algebra

To keep this article somewhat self-contained, this subsection tries to provide an overview or partial introduction to

geometric algebra, with many identities and formulas, aimed at helping the reader to more quickly see how quaternion

rotations can be used in geometric algebra. See the references and other works on geometric algebra for general

introductions to the subject.

Studying the algebra of rotations teaches a lot about the geometric algebra. In geometric algebra, the products and quotients

of vectors are also biradials or transition operators, and are very similar to quaternions. In fact, they are quaternions, but ina different form, with different terminology, and in a different, more generalized algebra. The quotient is still the

transition operator that takes to as . The transition of to is a composition of a rotation into the line

of followed by a scaling to the length of .

Quaternion vectors are imaginary directed quantities

A real quaternion has a pure quaternion or quaternion vector part

and a scalar part , where the scalars are limited to real numbers. The real

quaternions are simply the quaternions commonly used.

The quaternion vector units are used as a basis for representing points and directed quantities in a three-dimensional space. However, the square of a vector quantity in quaternions is . In effect, the directed

quantity is treated as a pure imaginary quantity, even though its components have been limited to real

numbers and the quantity should be considered real and have a real positive squared quantity. This is one of the problems

with quaternions and is part of the motivation for avoiding quaternion vector quotients and products in ordinary vector

calculus or analysis, as introduced in the 1901 book Vector Analysis[2] by E. B. Wilson that is founded upon the lectures of

J. W. Gibbs.

Euclidean vectors are real directed quantities

A different orthonormal basis of base vector units for three-dimensional space can be used instead of .

The directed quantity can be rewritten on this basis. If we now require that ,

then we also require that . The vector can now be called a Euclidean vector , and are

the Euclidean vector units of a three-dimensional Euclidean space that may be variously denoted or . The scalars

are still limited to real numbers.

Clifford geometric algebras

The algebra, variously denoted or , of the vector units is called the Clifford geometric algebra of

Euclidean -space, and is also denoted . We are only concerned with this particular Clifford geometric algebra in

all that follows here. This algebra differs from the quaternion geometric algebra of . However, the quaternion units

are isomorphic to the three unit -blades of this Clifford algebra, allowing quaternion concepts to be

used within the subalgebra of these three unit -blades.

Clifford algebras also allow more vector units than three, and also to have some units that square positive and some that

square negative. An example is the Clifford algebra having the vector units and ,

where the space is denoted and its algebra is denoted . The algebra is very useful and is called the conformal

geometric algebra of Euclidean physical space (C3GA). Its generalization to higher dimensions is called the homogeneous

model of [3](27). The discussion of C3GA is beyond the scope of this article, except to mention that quaternion rotation

concepts are central to using C3GA, wherein operators of the form generate all the ordinary transformations on

, including translations.

Clifford geometric algebra is also commonly called just geometric algebra. Geometric algebra, in the form discussed here,

is introduced in the 1984 book Clifford Algebra to Geometric Calculus, A Unified Language for Mathematics and

Physics[4] by authors David Hestenes and Garret Sobczyk. Geometric algebra derives from earlier work on the abstract

algebras of Clifford numbers that assign little or no geometrical meanings or interpretations. The emphasis of geometric

algebra is on geometrical interpretations and applications. Clifford geometric algebra is named after William Kingdon

Clifford. Clifford wrote papers, including Applications of Grassmann’s Extensive Algebra (1878)[5] and On The


d 61 15.4.2015 0:32


24/61

Classification of Geometric Algebras (1876)[6], that introduced geometric algebra as a unification of the earlier geometric

algebras of Argand, Möbius, Gauss, Peacock, Hamilton, Grassmann, Saint-Venant, Kirkman, Cauchy, and Peirce. Clifford

died in 1879 before he could fully develop and publish his ideas on geometric algebra.

Rules for the quaternion and Euclidean vector units

While the quaternion units have the rule or formula , the Clifford units

of have the formula

which is also showing some extra identities that map to quaternion quadrantal versors. The quaternion units map into as

the Euclidean vector biradials, which are called unit bivectors or -blades in Clifford algebra

and it is seen that, in terms of the quotients, the mapping is quite simple when we view the quaternion units as quadrantal

versor rotation axes being held in right-hand on right-handed axes model, or held in left-hand on left-handed axes model

while fingers curl round the rotational direction as thumb points in the positive direction of the axis. The inverse of a

quaternion unit is its negative or conjugate, while the inverse of a Euclidean unit is itself. The square of a unit bivector is

, as for example .

For , different units multiply anti-commutatively and are identical to their outer product

, where the outer product symbol between different base vector units is only synonymous or

symbolic of their perpendicularity, or of a angle between the units. The inner product symbol between same base

vector units is again only synonymous or symbolic of their parallelism, or of a zero angle between them. We also have the

important fundamental geometric product of vector units which is a true formula since either

and results in , or and results in .

The outer product

The formulas contain the product called the outer product . The outer product of vectors has the same magnitude as the

cross product of vectors, but the result is not a vector but is a value called a -vector or bivector . It will be shown that

the cross product of vectors is given by the formula . The outer product of parallel vector

components is zero , and the outer product of perpendicular vector components is an irreducible

symbolic outer product expression of the perpendicular vector components. The outer product of vectors is a new

-dimensional geometrical object representing a geometrical measure of their perpendicularity or space.

The appearance of the outer and inner product symbols between the different and same orthonormal base vector

units in the rule formulas, and the relation to the value and , can also serve to define these products in terms of simple

algebraic multiplications of the units in ways consistent with the rule formula. The product of algebraic multiplication of

two values in terms of only base vector units and scalars, obeying the basic rule formula, is known as the geometricproduct. The absence of any product symbol means algebraic multiplication consistent with rules, i.e. a geometric product.

An important result in the outer product of vectors is


d 61 15.4.2015 0:32


25/61

that one of the two parallel components is always arbitrarily scaled or deleted , such that or can be moved or

translated parallel to or (perhaps until intersecting or becoming collinear with another object or line of interest)

without affecting the result of their outer product. The value is the irreducible unit bivector of the outer product .

It will be shown that represents the plane containing and as a unit of directed and oriented area and that it also acts

as the quadrantal versor of the plane, similar to a unit vector in quaternions.

Outer products are geometrical values for imaginaries

The outer product expression can generally be written

where the determinant of the matrix of vectors in an -dimensional space is the -volume

of a -parallelotope formed by the -frame of the vectors . The outer product of the vector-units

can be regarded as a type of imaginary value or symbol, representing physical perpendicularity of

the unit vectors in up to physical dimensions, and representing imaginary perpendicularity in higher dimensions

. Similar to the imaginary value symbol , the imaginary product symbol is irreducible

to a real value by itself. However, the square of an imaginary product, such as , is a real value.

Generally, it is shown that

holds good for base unit vectors of an -dimensional Euclidean space where . Taking square roots, we might

(but should not) write and find that the right side has lost all of the vectors or

axes of the left side (a geometrical value), and that the right side is generally a kind of complex number (a metrical value).

However, this direct identification of a geometrical value to a metrical value is not valid . Nevertheless, it can be said that

the geometrical -parallelotope unit may represent a particular extension or instance of the

ordinary complex number symbol when , and of the split-complex or hyperbolic number

symbol when . Multivectors of the form , being a scalar plus

-blade, behave either as ordinary complex values on a complex plane or as hyperbolic values on a hyperbolic plane. The

unit -parallelotopes can be interpreted as geometrically specific objects that represent the metricalvalues and which by themselves represent unknown, lost, or arbitrary geometrical objects. Similar to

quaternions and losing an axis of rotation at angles and , it could be a problem to lose a -blade or -vector

used as a representation of a quaternion vector or axis of rotation.

Outer product is identified with scalar multiplication

The outer product with scalars is identical to ordinary scalar multiplication. Scalars are considered to be perpendicular to all

values, including to another scalar so that represents the -volume of an unknown

-parallelotope. It is not certain that is an area since an area is represented by , the outer product of two vectors. All

we can say about is that it is a scalar product, or pure inner product (this may seem to be a backwards statement), or

pure measure of parallelism between geometrical objects in the same dimensions by non-zero amounts, or the result of a

geometrical object contracted to zero dimensions by inner product with another geometrical object. Contraction is the

generalization of the vector dot product to dot products of higher-dimensional (or grade) geometrical objects represented by

outer product expressions. An outer product of scalars can be reduced to

, where and it represents an identity matrix that has lost its

vectors and has an indeterminate dimension, or has become dimensionless (non-vectorial) and allowed to be identical to the


d 61 15.4.2015 0:32


26/61

unit -vector .

As an additional comment, it should be noted that the formulas for the inner and outer products with scalars have not been

fully agreed upon by all mathematicians in geometric algebra[7](247). New or alternative formulas are beyond the scope of

this article. Most of the currently available software for geometric algebra continues to use the rules that the inner product

with a scalar is zero, and the outer product with a scalar is identical to ordinary scalar multiplication, and these are the rules

followed in this article.

A proposed change to the inner product says that the new inner product with a scalar is to be also identical to scalar

multiplication, giving the formulas[7](279)

where is any scalar, is any vector, and is any multivector. If the multivector contains a scalar component,

represented by the scalar grade selector , then subtracts one of the two copies. If , where is a

blade (no scalar component), then the “Chevalley formula”

continues to be valid. However, in general the Chevalley formula with any multivector is no longer valid under thischange. This proposed change is essentially the same as the (fat) dot product, but it tries not to introduce a new product

symbol.

Grades, blades, and vectors

The outer product expressions , , , and represent, respectively, the geometrical values or

objects: a -blade or metrical scalar in a -dimensional scalar subspace, a -vector or geometrical vector directed-

length in the -line, a -vector or geometrical bivector directed-area of parallelogram in the -plane (or -plane),

and a -vector or geometrical trivector directed-volume of parallelepiped in the -space (or -space).

A sum of linearly independent -blades is called a -vector , where the number is called the grade. The three vector units

are the unit -blades and their linear combinations are grade- vectors in a -dimensional subspace of . Thethree unit -blades are , , and , and a -vector has the general form

representing a pure quaternion vector or grade- vector in a -dimensional subspace of . The

single unit -blade , called the unit pseudoscalar , is a grade- vector in a -dimensional subspace of

. There is a total of -blade dimensions in .

The outer product of a -blade and -blade is a -blade, or if any vector is common between the blades multiplied

since they do not span any area or volume (no perpendicular component).

A multivector is a sum or linear combination of linearly independent -vectors of various grades .

There are no -blades or larger outer product expressions or objects than the 3-blade pseudoscalar in , since there are

only three base vector units .

Outer product expressions construct representations of directed geometric objects or values that are within a certain

attitude (a space spanned by a set of unit -vectors), have a positive scalar magnitude, and have a direction or orientation

(relative to the attitude or space) that is represented by a positive or negative sign on the magnitude.

The value of an outer product expression can also be represented by a matrix containing the vectors as elements where the

determinant gives the signed magnitude relative to the orientation of the basis.

The inner product

The product called the inner product is the complementary product to the outer product and is similar to the dot

product of vector analysis and linear algebra, except that the inner product of quaternion vectors, if represented by as pseudovectors, is the negative of the value given by the dot product in actual quaternions. The

Clifford algebra extends the inner product to take inner products of multivectors. The inner product of parallel

components is a scalar similar to multiplying numbers on a real or complex number line or axis, and the inner


d 61 15.4.2015 0:32


27/61

product of perpendicular components is zero . The inner product is a geometrical measure (a

multivector) of parallelism. Between two -vectors of the same grade, the inner product reduces to a pure metrical

measure (a scalar) of their parallelism, and no contracted (grade reduced) geometrical outer product expressions, objects,

or vector components will remain.

An important result in the inner product of vectors is

that one of the two perpendicular components is always arbitrarily scaled or deleted , such that or can be moved or

translated perpendicular to or without affecting the result of their inner product.

The inner product of a -blade and -blade is a -blade, which is often a -blade scalar. Scalars are considered to be

perpendicular to all values, even to other scalars; therefore, the inner product involving a scalar factor is zero. Inner

products that produce a scalar can be represented by the determinant of a matrix formed by the multiplication of two other

matrices representing outer products of equal grade.

A note on terminology may be appropriate here. The “inner” and “outer” products discussed in this article are those of the

Clifford geometric algebra. In the terminology of matrix linear algebra, an “inner product”, “dot product”, or “scalar

product” is an ordinary matrix multiplication of a row vector with a column vector that produces a single scalar element,

and is the means by which the dot product of vectors is computed in vector calculus. Also in the terminology of matrix

linear algebra, the “outer product” is the multiplication of a column vector with a row vector that produces a matrix. In the

terminology of Hermann Grassmann’s exterior algebra, there are “ interior” and “exterior” products, which are very closely

related to the Clifford geometric algebra’s inner and outer products. Geometric algebra has many distinct products to

represent scalar products, inner products, outer products, dot products, contraction products, and other specialized products

that can relate to products or concepts used in other algebras. Although distinct, the various products in geometric algebra

are all derived from, or are parts of, the geometric product .

The geometric product of two Euclidean vectors is the sum of their inner and outer products. This is similar to the

quaternion product of two pure quaternions (quaternion vectors), which is their cross product minus their dot product. The

quaternion product is the geometric product for quaternions. In Clifford geometric algebra, the geometric product

generalizes the quaternion product to more than four dimensions.

The sum of the inner and outer products

While the outer and inner products are of complementary magnitude according to the trigonometric functions, neither is a

complete product of vectors since each deletes a parallel or perpendicular component in one of the factors. Outer and inner

products are geometric measures of perpendicularity and parallelism, respectively. The complementary magnitudes suggest

that they are the projected components of yet some other value that is vectorially equal to their sum

This sum appears very similar to the quaternion vector biradial , and they can be shown to be

related as duals by dividing by the unit pseudoscalar. Considering unit vectors, and quaternion vectors only on the left of

and Euclidean vectors for the rest, we have

Since we started looking at the product first, the angle is now seen to represent the angle measured positive from

to , and goes the reverse way from to . Unit bivector represents a quadrantal versor in the direction from

to , or and . If we would agree to go back and reverse the sign on , we would have the usual

rotational orientation used in quaternion rotation biradials. However, unit bivectors like do in general represent the


d 61 15.4.2015 0:32


28/61

negative rotational direction compared to the similar looking cross product. For example, in quaternions is

the quadrantal rotation versor or axis that rotates from toward , while in Euclidean vectors the similar looking unit

bivector is a quadrantal versor, or geometric multiplier that effects planar rotation by , that rotates

from towards , according to

For unit vectors, the direct expression of this rotation is . We also could again agree to reverse

angles and then make as quadrantal versor from towards to match the Euclidean biradial where

the angle would be positive from to . The choice of orientation, to measure an angle positive or negative from to

or to , is the sign on the angle used in the sine function of the outer product and this is why

holds in general. The choice of orientation could also be viewed as a choice between axis or inverse

axis, or between angle or reverse angle.

When the angle between vectors and , the product or sum reduces to justthe positive scalar which is the expected product of parallel vectors; in quaternions, this product would be

negative since quaternion vectors square negative and quaternions treat vectors as imaginary directed quantities in a space

that is spanned by three imaginary units. Here, we have a real Euclidean space spanned by real units.

When , the sum reduces to which is a directed-area parallelogram of magnitude in the plane of

unit bivector , and so the product represents the expected geometrical object.

For other values of , the result is a multivector value where the scalar part represents the projection of the multivector

onto a scalar axis, and the bivector part represents the projection of the multivector onto a bivector axis.

Since is a unit bivector, it squares to in the same way as a quaternion unit vector; and therefore, the sum of the inner

and outer products of two Euclidean vectors has the form and characteristics of an ordinary complex number or aquaternion. The magnitude of this number is found by taking the square root of the product of conjugates, by which we can

see that the magnitude really is . Also by similarity to complex numbers or quaternions, we know this must be the

product from which the magnitudes and angle are taken. Yet, the Euclidean vectors and themselves do not have

characteristics of ordinary complex numbers. Since and square positive, they do have the characteristics of split-

complex numbers or hyperbolic numbers.

The multivector can be written as

which shows more properties of the algebra, and is the reflection of the scalar and bivector components of the product

in the vector . Each vector in the -blade is rotated or reflected individually, and that is the general procedure to

rotate blades.

The reverse of a blade

The outer product, that constructed the bivector , has the characteristic of constructing unit blades that square or

in a certain pattern. In general, it can be shown by a sequence of anti-commutative identities, that any blade and its reverse

are related by


d 61 15.4.2015 0:32


29/61

Triangular number pyramid.

where is the grade of the blade. The square of any -blade has the sign given by . For

, the pattern looks like so that at for , then

.

The value can represent a triangular area that is equal to the total number of bivector reverses. For example,

to reverse the for , it takes bivector reverses to move to the right side

and it takes bivector reverses, respectively, to move into their reversed spots. The last vector

takes zero reverses. A triangle of these bivector reverse counts can be constructed as the triangle of numbers

that has a total areal count of the base times the height

which is the triangular number (sum of integers) formula for .

The sum of integers is proved algebraically by adding the numbers counting up, plus adding the numbers counting down as

follows.

The square of a blade

In geometric algebra, squaring a blade is an important algebraic operation. It extends the concept of squaring a vector to

more dimensions or to higher grades of vectors.

Consider vectors written in component form as

Using the formula for the reverse of a -blade


d 61 15.4.2015 0:32


30/61

we can write the square of a -blade as

Using the recursive formula for the inner product of blades, we can expand this inner product as

To demonstrate this result for , let’s take

and factoring this determinant gives


d 61 15.4.2015 0:32


31/61

The square of a -blade has a sign given by , and has a positive magnitude, or absolute value, given by the

square of the determinant of the scalar coordinates.

The determinant of the scalar coordinates

is the signed, or oriented, -volume of the directed -parallelotope framed by the vectors. Multiplying this determinant

by gives the signed -volume of the directed -simplex framed by the vectors.

It can also be seen that the product of two parallel -blades, on the same unit -blade or in the same -space, is the scalar

product


d 61 15.4.2015 0:32


32/61

The sign factor is eliminated by reversing rows in the first determinant.

More generally, the inner product of two -blades and of vectors in a -dimensional base space or -space spanned

by orthonormal units is the determinant

If then this determinant can be factored into the product of two determinants, but otherwise it cannot since the

factored matrices would be and which are not square and have no determinants.

The free software SymPy (http://sympy.org), for symbolic mathematics using python, includes a Geometric Algebra Module

and interactive calculator console isympy. The isympy console can be used demonstrate these results. A simple example of

console interaction follows to compute the product of parallel -blades


d 61 15.4.2015 0:32


33/61

$isympy

>>> from sympy.galgebra.ga import *

>>> (e1,e2,e3) = MV.setup('e*1|2|3',metric='[1,1,1]')

>>> (a1,a2,a3) = symbols('a1 a2 a3')

>>> (b1,b2,b3) = symbols('b1 b2 b3')

>>> a1 = 3*e1 + 4*e2 + 5*e3

>>> a2 = 2*e1 + 4*e2 + 5*e3

>>> a3 = 9*e1 + 6*e2 + 9*e3

>>> b1 = 9*e1 + 2*e2 + 3*e3

>>> b2 = 6*e1 + 5*e2 + 8*e3

>>> b3 = 2*e1 + 4*e2 + 7*e3

>>> (a1^a2^a3)|(b1^b2^b3)

-102

>>> (A,B) = symbols('A B')

>>> A = Matrix([[9,6,9],[2,4,5],[3,4,5]])

>>> B = Matrix([[9,2,3],[6,5,8],[2,4,7]])>>> A.det() * B.det()

-102

The geometric product ab of Euclidean vectors

For Euclidean vectors and , we have the products

which are called the geometric product of vectors. The notation and concept of the quaternion conjugate operation can

be usefully adapted to the Euclidean biradials, such that where there was a pure quaternion unit vector part to take

negative (as axis or angle) there is now a similar unit bivector part . The unit bivector is now being taken as

quadrantal versor from toward , and angle is positive from toward .

The geometric product can also be shown to hold for any product of a vector and multivector, such as

and , although the evaluation of these inner and outer products of a

vector and multivector require usage of the formulas for expansion of inner and outer products. The geometric product

usually has no special symbol, and the product is invoked, as shown, by spaces and juxtaposition of the values to be

multiplied.


d 61 15.4.2015 0:32


34/61

The quaternion product PQ of even-grade multivectors

The geometric product is found to be the same as the quaternion product when multiplying values that isomorphically

represent quaternions in . The quaternion representation and multiplications are

where the pseudoscalar multiplies commutatively (like a scalar) with all other values, and , causing some

changes in signs.

Dual and undual to convert to and from quaternions

Taking the dual (dividing by on the right) of the Euclidean vector as produces its pure quaternion vector

representation . Taking the undual (multiplying by on the right) of the bivector as reproduces the

Euclidean vector represented by the pure quaternion vector .

The inner product of quaternionic bivectors has negative sign

In the equation that represents quaternion multiplication , the sign on is positive, making the equation appear

different here than in the actual quaternion formula for quaternion multiplication that represents. Therefore,

, and is the correct dot product of the undual vectors represented. This difference in sign is an

issue to be aware of, but it is technically not a flaw or problem. The other values, and the values taken by grades

and , are completely consistent with quaternion multiplication.

The commutator product

The product symbol is being used in two different contexts or meanings. In it is the Euclidean vector cross

product and is the dual of the vector outer product. In it is called the commutator product , and is representing a

quaternion vector cross product which is then taken undual to be the Euclidean vector cross product result. The

commutator product is defined generally, in terms of any multivectors, by the identity

where we see this form for . For vectors, we happen to have . Writing


d 61 15.4.2015 0:32


35/61

is not the commutator product; it is the vector cross product or dual of the vector outer product.

The meaning of the symbol ( ) is overloaded in the context of vectors. While outer products are associative

, commutator products are non-associative and obey the Jacobi identity

. This non-associative algebra of commutator products is mostly

avoided, but there is a formula

with the conditions and , that relates multiplying bivectors and multivectors that contain no

-vector components . However, this conditional formula is of little interest for our needs.

We are only concerned with the representation of quaternions as bivectors, and the formulas of interest are the ones

representing quotients and products of quaternion unit vectors, namely the biradials or versors for rotation from vector to

where duals and take the Euclidean vectors into quaternion representations and . An undual

takes a quaternion vector representation back to a Euclidean vector .

Note well, that the inner product of two quaternion vectors represented as bivectors

is the negative of a dot product, or inner product of two Euclidean vectors. Two quaternion vectors multiply as

and two quaternion vectors represented as bivectors multiply as

This difference in sign, on the inner product and on dot product, must be paid careful attention.

Evaluating expressions in Clifford geometric algebra

The products of Clifford geometric algebra are somewhat complicated, and only careful usage of formulas and identities

will give a correct evaluation. The geometric product is often written as the sum of the inner and outer products, but that

formula only applies generally when at least one of the values being multiplied is a vector. Identities are used to change a

product that doesn’t involve a vector into one that does, and these identities are used recursively until the product is


d 61 15.4.2015 0:32


36/61

expanded enough to complete the product evaluation using other identities for products of vectors and multivectors.

For example, the inner product evaluates almost as if are merely scalars once

they are arranged on the right side by using the anti-commutative multiplication identity of perpendicular vectors; the result

is . The evaluation of the geometric product might at first be thought to be

or , or even , but

it is none of them since neither operand of this geometric product is a vector. The correct evaluation using a recursive

expansion identity (see identities below) is

This geometric product evaluation was made more complicated by it not being a simpler product of a vector with a

multivector. The evaluation of is

which is the correct result that matches the rule formula above. The rules or formulas applied here are listed in the identities

below. Of course, these products can also be evaluated more simply by using the identities already above and applying the

rule for perpendicular vectors as needed to move values to the left or right in order to square vectors into

scalars. The evaluation then looks like this

The general form of a multivector A

While looking below at some of the fundamental identities of this Clifford algebra , it may be helpful to follow along

with the following additional explanations or interpretations of the values in the identities. Additional reading in the

literature on Clifford geometric algebra is required to more fully understand all the concepts.

The value

is the general form of a value, called a multivector , in . The multivector has parts called the -vector part [8](41)

of , similar to taking the scalar and vector parts of a quaternion except that there are four different linearly

independent parts. The number itself is called the grade, and is the grade part operator or selector on . The

values in a part are also said to be of grade k or have grade k [4](3).

The part

is a real number scalar, sometimes called a 0-vector or 0-blade, and is often associated with the cosine of an angle as in

quaternions.

The part


d 61 15.4.2015 0:32


37/61

with real scalars , is a Euclidean vector, called a 1-vector or just a vector , and its three linearly independent

orthogonal components are called 1-blades.

The part

with real scalars , is called a bivector (not the bivector in biquaternions) or 2-vector , and its three linearly

independent orthogonal components are called 2-blades, and this part also represents a pure

quaternion vector.

The part

with real scalar , is called a 3-vector or 3-blade, or a pseudoscalar .

There are a total of linearly independent components or blades in all the parts of .

In a Clifford geometric algebra with orthonormal base vector units, there are unique blades, and therefore the

algebra has values in a -dimensional real number space. One way to see how there are unique blades is by

considering the expression

where the act as presence bits. The corresponding binary number has possible values.

When all presence bits are zero, the resulting number is the grade-0 base unit. The grade of a blade is the sum of the

presence bits. If the ordering of the units is changed in this expression, then by anti-commutativity there would only be

differences in signs but no additional unique blades. The space of the blades is called the Grassmann space

generated by the vector space .

The algebra of the even-grade parts is quaternions

The algebra of multivectors having only the even grade parts is a closed algebra of values and

isomorphic to quaternions in general, so it is a subalgebra of . The algebra of multivectors having only the odd grade

parts is not closed since their products or quotients are in the even grade parts, which can be

thought of again as quaternions as just mentioned. The product of an even part and odd part is an odd part. This relationship

between odd and even parts can be notated as , and , and ,

and . If the odd parts are taken to represent only pure quaternions (vectors),

letting , and the even parts are taken to represent only impure quaternions (quaternions

with non-zero scalar part), then we will find that can (abstractly) represent the quaternion versor

operator , which will take the Clifford algebra form .

In this form, is a unit bivector that directly represents a plane of rotation, or represents a quaternion unit vector axis of

rotation , and it is the quadrantal versor of the plane represented by . The value of is still, in terms of

quaternion biradials, the product or quotient of two perpendicular unit vectors. However, is now given as the product or

quotient of two perpendicular Euclidean unit 1-vectors , or as the product or quotient of two perpendicular

quaternion unit vectors represented by unit 2-vectors

or as the dual of a unit vector rotational axis . More generally, by allowing and to again have any angleseparating them, and where we had the biradial in quaternions, we now have the unit multivector of even grade


d 61 15.4.2015 0:32


38/61

representing the same versor, and

is the unit bivector of the plane of the Euclidean unit vectors to . operates as quadrantal versor directly on

Euclidean vectors in the plane of , such that is rotated by in the plane and in the direction

toward .

The reversor ,

is the reverse[8](45) of denoted .

The quadrantal reversor is

Representing a quaternion versor using an odd-grade part

It is also possible for the odd part to represent a quaternion if it takes the form

where the Euclidean unit vector is the rotational axis, and is the rotational angle according to a right-hand or left-hand

rule on right-handed or left-handed axes. A Euclidean vector can then be rotated by round as

The rotation (b/a)v(a/b) in Euclidean vectors

In general, the operation , called taking the dual of , transforms a Euclidean planar-polar vector into an

orthogonal -vector in the dual space that represents the same vector but as a quaternion axial-pole vector (also called a

pseudovector ) that can operate as a quadrantal versor if it has unit scale. The operation can be called taking theundual of . By taking the dual of Euclidean vectors to transform them into quaternions, ordinary quaternion rotation,

in terms of biradials, can be performed and then the undual can be used to transform the rotated result back to a Euclidean

vector. Assuming unit vectors , then rotation of by twice the angle from to is

which shows that the products or quotients of Euclidean unit vectors represent quaternion versors that can directly operate

on a Euclidean vector to perform a rotation in the Euclidean vector space. Therefore, it is not required to take duals totransform to quaternion representations of rotations, nor then undual any rotated results.

The following are some of the most important identities for the Clifford geometric algebra of Euclidean 3-space .


d 61 15.4.2015 0:32


39/61

Identities: Euclidean base unit vectors

Euclidean base vector units in :

Identities: Euclidean 1-vectors

Euclidean -blades and -vector (vector) in :

Identities: The geometric product ba of 1-vectors

The products of vectors in terms of the units. The geometric, inner, and outer products are evaluated directly in terms of the

base units, and verifies that the geometric product of vectors is the sum of the inner and outer products of vectors. The

outer product is found to be equal to the negative of the commutator product of their duals, which represents the negative

cross product of their quaternion representations. The product is shown as a scaled versor that rotates from to :


d 61 15.4.2015 0:32


40/61

Identities: The quaternion unit 2-vectors

Isomorphic (1 to 1) map between unit -blades and quaternion units, where the dual of a planar-polar Euclidean

vector still represents but as an axial-pole quaternion vector , which is a quadrantal versor of the plane

perpendicular to if :

It is also important to show that the inner products of the quaternion mapping give the correct results

ternion rotation biradial - Wikipedia, the free encyclopedia https://en.wikipedia.org/w/index.php?title=Quaternio

Quaternion Rotation Biradial - Wikipedia, The Free Encyclopedia20150415003245

Documents

Transcript of Quaternion Rotation Biradial - Wikipedia, The Free Encyclopedia20150415003245